How many bits are used for each colour channel internally by DXO Photo Lab?

DxO PhotoLab
41

do not play with definitions … you are not a toddler in a sandbox anymore

and what “color” exactly is a number written by a camera’s firmware for a specific sensel in your raw file :slight_smile: ?

please finally learn color 101 - input devices DO NOT HAVE a GAMUT by defintion ( I know you don’t like definitions ) - output devices do or properly defined color spaces …

Also to capture something does not give you any one true idea what you captured - you can only arbitrary interpret that ( and interpretations can vary )

for example if I illuminate a sensel with a light of X nm wavelength (assume this will go through) for a while and then with a light of Y nm wavelength (assume this will go through) for another while and then readout will be a digital number Z - what did you capture :slight_smile: ? at no point there was X+Y nm light in play …

as I noted - try to compehend IR + UV range capture by a sensor for example… what color we are talking about exactly here ( or there ) :smiley: ?

here we go again sensors do not provide colors ( but you are beyond repair )… arbitrary (sic ! because - like it or not - there was no one true color transform yet invented for any “camera” by anybody, not to mention a lot of other factors like lenses for example ) color transforms executed by raw converters do result in “colors” appearing after that - so wider gamut of color space into which we decide to assign the data certainly helps vs a smaller gamut if we decide that we want to map something (that is not yet a color) to a color ( coordinates in a proper color space ) outside of a smaller gamut …

42

except of course you need to pay attention how raw converter works from start to finish … the mere fact that something was not clipped in raw channel does not mean that raw converter will not clip that in its internal pipeline… push/push of so called “exposure” sliders or similar controls can result in color shifts based on how color transform was implemented ( like with hue twists or not ) … to get as much light as possible is a nice concept until you are derailed by a practical implementations of the full pipeline if you try to push too hard - may or may not happen … I remember myself being an idiot ~10+ years ago ( in 2012 being precise ) and gelling my flash light to have shift it to magenta ( with https://mpex.com/rosco-4730-calcolor-30-magenta-20x24-sheet.html ) … somebody said - you saturate all raw channels properly and then just do a WB ( that was instead of putting a magenta filter on a lens )… well till I tried to use shipped camera profiles :slight_smile: … WB sliders range were not an issue, but guano was the result because well camera profiles did not work well at all … but I get a better saturation in “blue” and “red” channels in raw files indeed

raw channels from WB target illuminated by gelled flash - good stuff, no ?.. blue channel saturated as good as green, red just ~1/2 EV below … except I was not able to do a good raw conversion for a while instead of getting next day … but that was on me - trusted a concept of giving it as full saturation as possible and did not check that existing pipeline will cope out of the box

_1020060-Sel-2559-1406-366x236
_1020060-Sel-2559-1406-366x2361842×1992 92.5 KB

43

Highlight clipping whether it happens in the converter or at the A/D converter is highlight clipping.

I’m not gonna get into a bickering argument about highlight clipping like what’s also going on here about whether there is color in a raw file.

44

except you can work around the converter - by at least using a different one… clipping in raw channels in raw file data where you really need proper information ( not like specular highlights ) you can’t work around properly ( well may be generative AI can save some spots of course nowadays - but then why bother with ETTR ? )

45

Color: light with a certain wavelength.
gamut: a collection of colors->a collection of light with a certain wavelength. Produced by the outputdevice.
sensor: a device that is sensitive to light within a range of wavelength. Input device.
color filter array: a filter in front of the sensor, is filtering the light for a certain smaller range of wavelength. Input device.

Both the input device as the output device do have a range of wavelength they’re sensible to. A range of wavelength is gamut.
RAW file: contains the by the sensor collected analogue values in a binary value using a certain range or bitdepth.
RGB file: contains pixels created in the demosaicing process existing out of 3 colors. Is send to the output device to create an analogue output we can see with our eyes.

An image file doesn’t have a color. Colors are produced by the output device.

There’s something as a gamut for input devices.

George

1 Like
46

Well folks, let’s make it simple:

A file (read from a digicam) is just a bunch of bits. It neither contains an image nor colour…but it contains information about an image and its colours. But again, it’s just a bunch of bits.

:japanese_ogre:

1 Like
47

And that bunch of bits contains the digitized analogue values the sensor produced based on the range of light(gamut) the sensor received through the color filter array. The RAW file contains that information on the received input. A RGB file contains information of the desired output.

George

48

First you don’t know me. And we could talk about lot of definitions.

Second your tone exceeds the limit of acceptability.

Third, stop doing your professor about the very obvious things. I think most people on this forum know the basics of colors.

If you don’t understand what I told and are as narrow-minded as that, please, don’t respond me again.

49

You guys are in violent agreement in principle. You’re only arguing over semantics. Only arguing over the words.

It’s a stupid argument frankly.

50

This is what I meant.
Go back and follow the thread.

51

one more time - for dumb people , here it is explained in plain english = A Digital Camera Does Not Have A Color Gamut – Color-Image.com

thanks to " About the Author

Parker Plaisted is the author of this site. Parker has a B.S. degree in Physics-engineering from Washington and Lee University, an M.S. degree in Imaging Science from the Rochester Institute of Technology (RIT), and an M.B.A. degree with concentrations in marketing and business strategy from Vanderbilt University. While at RIT, Parker studied color science in the Munsell Color Science Laboratory under professors Mark D. Fairchild and Roy S. Berns. His classmates included Ricardo J. Motta, Mark E. Gorzynski, Mitchell R. Rosen, Taek Kim, Jason Peterson, Ranjit Bhaskar, Tom Orino, David Telep, and David Erdtmann. Parker’s primary thesis advisor was Dr. Ed Granger.

From 1994 to 1996, Parker was the Director of the Imaging Division at the RIT Research Corporation where he worked on color imaging projects with J. A. Stephen Viggiano, Milt Pearson, Nathan Moroney, David Brydges, Chris Sawran, Chris Pane, Jennifer Greenwald, Bill Hoagland, and Jeff Harman. His clients included Xerox Corporation, Hewlett-Packard Company, Eastman Kodak Company, Kimberly-Clark, CalComp, IBM Printing Systems, and Heidelberger Druckmaschinen.

Parker has made technical contributions to the development of color imaging systems and to software applications that create and use ICC profiles. He has more than 10 years of experience working on the development and implementation of color imaging systems.

Additional information is provided in the LinkedIn.com profile for Parker Plaisted and Google Scholar Profile for Parker Plaisted."

recoding something does not make that data a color …

52

depends on the formal definition of the “image” … things like gamut are clearly defined in color science and it is not applicable to input devices in principle … it is as idiotic as “high ISO noise” and “exposure triangle”

53

nope… it is not … The definition of color specifies that it is an attribute of visual sensation. If there is no visual sensation, then there is no color (c) … that is why some coordinates in some color spaces are NOT colors btw

54

nope, it is not - see above referenced explanation by somebody who is actually a color scientist :slight_smile:

55

“Color gamut: Range of colors produced by a coloration system.” To be a little clearer about this, the concept of a color gamut applies to systems that produce color (e.g., color printer, color television, color monitor, or color projector).

I think it’s fine if he wants to limit the concept of gamut to color-creating devices. It will not stop me from comparing these in any way to color measuring devices.
Color measuring: light->CFA->sensor->analogue value->digital value.
Color creating device:digital value->analogue value->led->light(color)
The big difference is tat the color creating device is making colors by putting 3 colors close to each other so we ‘see’ a unique mixed color.

From wikipedia Gamut - Wikipedia

CIExy1931_srgb_gamut
CIExy1931_srgb_gamut542×599 94 KB

What do you think those numbers on the side line of the horse shoe means? Well, surprising wavelength.

Gamuts are commonly represented as areas in the CIE 1931 chromaticity diagram, with the curved edge representing the monochromatic (single-wavelength) or spectral colors.

George

56

please refer to a proper definition of the gamut and why input devices do not have it … that’s about it …

by all means , go ahead

just in case if you do not know spectrophotometers do not have CFAs, but we will not stop you from dangling your ignorance … and digital value is not a coordinate in any color space unless an arbitrary color transform is executed ( and there is no single correct one… just so you know companies like Gretag MacBeth and X-Rite used different ones in their spectrophotometers :slight_smile: - because there was no single correct one and never will be )

57

it is not me - it is what color scientists actually do, unlike you … but feel free to invent you own “science”

58

Angry? Did your mother punish you?

George

59

Come along children. Settle down.

2 Likes